Abstract: In response to the problems of excessive computational complexity and poor accuracy in traditional maximum likelihood direction of arrival (ML-DOA) estimation, improved monarch butterfly optimization algorithm is proposed and applied to ML-DOA estimation. By using elite reverse learning strategies to optimize the initial monarch butterfly population and obtaining individuals with better fitness values, the diversity of the monarch butterfly population can be improved; In order to improve the individual optimization method, the mutation operation and adaptive strategy inspired by differential evolution are used for expanding the search scope of the algorithm; At the same time, the Gaussian-Cauchy mutation operator is introduced to adjust the mutation step size, preventing the algorithm from trapping in local optima, and apply IMBO to ML-DOA. Experiments show that the proposed algorithm has better convergence performance, lower root-mean-square deviation and less computation under different number of sources, SNR and population, compared with the conventional DOA estimation algorithm.

Key words: direction of arrival, maximum likelihood estimation, butterfly optimization algorithm, elite reverse learning, adaptive strategy, mutation operator